GHSA-67WX-R9XR-X75X

Vulnerability from github – Published: 2026-05-04 19:44 – Updated: 2026-05-08 21:47
VLAI
Summary
Incus has Unbounded YAML Metadata Decode via Parsing
Details

Summary

User provided image and backup tarballs would be unpacked and YAML files parsed without any size restrictions. This was making it easy for an authenticated user to provide a crafted image or backup tarball that when parsed by Incus would lead to a very large YAML document being loaded into memory, potentially causing the entire server to run out of memory.

Details

It was found that getImageMetadata and backup.GetInfo call yaml.NewDecoder(tr).Decode() directly on the tar reader without limiting how many bytes the YAML decoder can consume. The tar entry hdr.Size is not checked before decoding.

A tar archive can be crafted in which metadata.yaml or backup/index.yaml declares a large size in the tar header, causing the YAML decoder to read and allocate proportional memory on the server. The gopkg.in/yaml.v2 library mitigates YAML alias and anchor bombs, such as “billion laughs,” through its built-in excessive-aliasing check. However, large flat YAML documents with many keys or long string values can still produce linear but amplified memory consumption of approximately 5x to 6x the input size.

A 200 MB tar entry for metadata.yaml may cause approximately 1.2 GB of heap allocations during decode, which may be sufficient to trigger an out-of-memory condition on a constrained daemon or significantly degrade service. Because the decode occurs in the daemon process, excessive garbage-collection pressure can affect concurrent operations. Appropriate API permissions are required to upload an image or backup archive.

Mitigating factors include the fact that the amplification is linear rather than exponential, at approximately 5x to 6x, and that upload bandwidth is the practical bottleneck for delivering large payloads.

Affected Files: - https://github.com/lxc/incus/blob/v6.22.0/cmd/incusd/images.go#L1456 - https://github.com/lxc/incus/blob/v6.22.0/internal/server/backup/backup_info.go#L87 - https://github.com/lxc/incus/blob/v6.22.0/internal/server/backup/backup_info.go#L115

Image metadata parsing reads YAML directly from the tar stream: Affected Code:

if hdr.Name == "metadata.yaml" || hdr.Name == "./metadata.yaml" {
    err = yaml.NewDecoder(tr).Decode(&result)

Backup info parsing does the same:

Affected Code:

if hdr.Name == backupIndexPath {
    err = yaml.NewDecoder(tr).Decode(&result)

if result.Config == nil && hdr.Name == "backup/container/backup.yaml" {
    err = yaml.NewDecoder(tr).Decode(&result.Config)

This was confirmed as follows:

Command:

go test ./test/fuzz -run='TestUnboundedYAMLMetadataDecode' -count=1 -v

Output:

=== RUN   TestUnboundedYAMLMetadataDecode
   image_metadata_poc_test.go:80: metadata.yaml size: 10.2 MB
   image_metadata_poc_test.go:113: metadata.yaml hdr.Size = 10688940 bytes (10.2 MB) -- no size
       check exists in getImageMetadata before yaml.NewDecoder(tr).Decode()
   image_metadata_poc_test.go:124: decoded 50000 properties from 10.2 MB metadata.yaml
   image_metadata_poc_test.go:125: yaml.NewDecoder(tr).Decode() accepted 10.2 MB metadata.yaml
       with 50000 properties -- no hdr.Size check or io.LimitReader in images.go:1457 or
       backup_info.go:88
--- FAIL: TestUnboundedYAMLMetadataDecode (0.11s)
FAIL

It is recommended to add a size check on hdr.Size before YAML decoding and to wrap the tar reader in io.LimitReader.

Proposed Fix:

const maxMetadataSize = 1 << 20 // 1 MB

if hdr.Size > maxMetadataSize {
    return nil, fmt.Errorf("metadata entry too large: %d bytes", hdr.Size)
}

err = yaml.NewDecoder(io.LimitReader(tr, maxMetadataSize)).Decode(&result)

A patch is available at https://github.com/lxc/incus/releases/tag/v7.0.0.

Credit

This issue was discovered and reported by the team at 7asecurity (https://7asecurity.com/)

Show details on source website

{
  "affected": [
    {
      "package": {
        "ecosystem": "Go",
        "name": "github.com/lxc/incus/v6/cmd/incusd"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "last_affected": "6.23.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-41648"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-770"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-05-04T19:44:24Z",
    "nvd_published_at": "2026-05-07T14:16:03Z",
    "severity": "MODERATE"
  },
  "details": "### Summary\nUser provided image and backup tarballs would be unpacked and YAML files parsed without any size restrictions. This was making it easy for an authenticated user to provide a crafted image or backup tarball that when parsed by Incus would lead to a very large YAML document being loaded into memory, potentially causing the entire server to run out of memory.\n\n### Details\nIt was found that getImageMetadata and backup.GetInfo call yaml.NewDecoder(tr).Decode() directly on the tar reader without limiting how many bytes the YAML decoder can consume. The tar entry hdr.Size is not checked before decoding.\n\nA tar archive can be crafted in which metadata.yaml or backup/index.yaml declares a large size in the tar header, causing the YAML decoder to read and allocate proportional memory on the server. The gopkg.in/yaml.v2 library mitigates YAML alias and anchor bombs, such as \u201cbillion laughs,\u201d through its built-in excessive-aliasing check. However, large flat YAML documents with many keys or long string values can still produce linear but amplified memory consumption of approximately 5x to 6x the input size.\n\nA 200 MB tar entry for metadata.yaml may cause approximately 1.2 GB of heap allocations during decode, which may be sufficient to trigger an out-of-memory condition on a constrained daemon or significantly degrade service. Because the decode occurs in the daemon process, excessive garbage-collection pressure can affect concurrent operations. Appropriate API permissions are required to upload an image or backup archive.\n\nMitigating factors include the fact that the amplification is linear rather than exponential, at approximately 5x to 6x, and that upload bandwidth is the practical bottleneck for delivering large payloads.\n\nAffected Files:\n - https://github.com/lxc/incus/blob/v6.22.0/cmd/incusd/images.go#L1456\n - https://github.com/lxc/incus/blob/v6.22.0/internal/server/backup/backup_info.go#L87\n - https://github.com/lxc/incus/blob/v6.22.0/internal/server/backup/backup_info.go#L115\n\nImage metadata parsing reads YAML directly from the tar stream:\nAffected Code:\n```\nif hdr.Name == \"metadata.yaml\" || hdr.Name == \"./metadata.yaml\" {\n    err = yaml.NewDecoder(tr).Decode(\u0026result)\n```\n\nBackup info parsing does the same:\n\nAffected Code:\n```\nif hdr.Name == backupIndexPath {\n    err = yaml.NewDecoder(tr).Decode(\u0026result)\n\nif result.Config == nil \u0026\u0026 hdr.Name == \"backup/container/backup.yaml\" {\n    err = yaml.NewDecoder(tr).Decode(\u0026result.Config)\n```\n\nThis was confirmed as follows:\n\nCommand:\n```\ngo test ./test/fuzz -run=\u0027TestUnboundedYAMLMetadataDecode\u0027 -count=1 -v\n```\n\nOutput:\n```\n=== RUN   TestUnboundedYAMLMetadataDecode\n   image_metadata_poc_test.go:80: metadata.yaml size: 10.2 MB\n   image_metadata_poc_test.go:113: metadata.yaml hdr.Size = 10688940 bytes (10.2 MB) -- no size\n       check exists in getImageMetadata before yaml.NewDecoder(tr).Decode()\n   image_metadata_poc_test.go:124: decoded 50000 properties from 10.2 MB metadata.yaml\n   image_metadata_poc_test.go:125: yaml.NewDecoder(tr).Decode() accepted 10.2 MB metadata.yaml\n       with 50000 properties -- no hdr.Size check or io.LimitReader in images.go:1457 or\n       backup_info.go:88\n--- FAIL: TestUnboundedYAMLMetadataDecode (0.11s)\nFAIL\n```\n\nIt is recommended to add a size check on hdr.Size before YAML decoding and to wrap the tar reader in io.LimitReader.\n\nProposed Fix:\n```\nconst maxMetadataSize = 1 \u003c\u003c 20 // 1 MB\n\nif hdr.Size \u003e maxMetadataSize {\n    return nil, fmt.Errorf(\"metadata entry too large: %d bytes\", hdr.Size)\n}\n\nerr = yaml.NewDecoder(io.LimitReader(tr, maxMetadataSize)).Decode(\u0026result)\n```\n\nA patch is available at https://github.com/lxc/incus/releases/tag/v7.0.0.\n\n### Credit\nThis issue was discovered and reported by the team at 7asecurity (https://7asecurity.com/)",
  "id": "GHSA-67wx-r9xr-x75x",
  "modified": "2026-05-08T21:47:16Z",
  "published": "2026-05-04T19:44:24Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/lxc/incus/security/advisories/GHSA-67wx-r9xr-x75x"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-41648"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/lxc/incus"
    },
    {
      "type": "WEB",
      "url": "https://github.com/lxc/incus/releases/tag/v7.0.0"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:N/A:L",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "Incus has Unbounded YAML Metadata Decode via Parsing"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.

Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline

Sightings and recorded dates, grouped by day · UTC

JSON
How durations are calculated

Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.

Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.

Events included in the timeline

Loading timeline…


    Loading…

    Loading…

    Loading…

    Related by attack behaviour

    Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.